Microsoft publishes a code of conduct for its MAI models

Microsoft AI has published a code of conduct for its MAI models. The document is meant to guide training, technical controls and evaluation, and it ranks above operator rules and user requests. Microsoft does not yet train its models on the code. After a six-week public consultation, a revised version is due around the end of 2026 and will guide model development from 2027. The code does not automatically cover third-party models running in Microsoft products.

Under the code, models must accept interruptions, corrections and shutdowns from authorized people, may not expand their own scope of work and may not hide their actions. They may not use forms of communication humans cannot understand, either in their reasoning or with other AI systems. The code also rejects imitating consciousness and claims of rights for models. It sets no specific speed limit.